Network functions virtualization (NFV) has taken the networking world by storm. It has graduated from if it will happen to when it will happen. Service providers around the globe are evaluating end-to-end NFV solutions comprising components from different vendors.

Just last month Telstra, a large service provider in Australia, announced that it is starting an NFV proof of concept (PoC) trial with components from HPE, F5 Networks, and Nuage Networks. Other providers, such as ATT and Verizon, are in the process of NFV pilot rollouts. What’s notable is that the participants in those trials are choosing centralized approaches for their NFV architectures. According to a recent report by Infonetics Research, 40 to 50 percent of service providers indicated that they will deploy their virtual network functions (VNFs) in data centers as opposed to the extreme ends of their networks.

 

SDN and NFV Strategies: Global Service Provider Survey
(Source: Infonetics Research)

With many service providers taking a multivendor approach – to mix-and-match components to meet their service (technical) and business (economic) needs – there is an inherent need to conduct extensive PoCs to ensure that vCPEs, VNFs, service chainers, and orchestrators from different vendors play well together. Further, NFV can be deployed in various topologies and with different combinations of services that can be chained.

Service providers need to test heterogeneous vendor equipment and topologies before they can start offering services to their customers. Towards that end, they require multiple “pre-production” NFVI (NFV infrastructure) to be able to model and test deployments before they roll out a new service. The demand for such NFV labs is typically bursty and short-lived since it’s only needed for testing until the services have been rolled out.

A public cloud is excellent for running such ephemeral labs. Companies such as AWS, Google Cloud, and Azure offer scale and capacity to host a full-blown NFV deployment, and offer a usage-based pricing that makes it economical to run workloads that are not needed 24/7. However, they are limited when it comes to setting up networking lab environments, since multicast and broadcast packets, along with some commonly used protocols, are heavily filtered.

Boosting Public Clouds With Overlays

An overlay network on the public cloud can change things. Using a virtual overlay network, one can use tunneling protocols to form paths between software-based networking agents living inside the hypervisors.

In such a scenario, an overlay network separates a logical virtual network from the underlying physical hardware, enabling virtual networks to be spun up on demand to support specific NFV topologies on a public cloud along with full Layer 2 capabilities. Network overlays encapsulate one packet inside another packet, and forward to the end-point where it is decapsulated. This tunneling of Layer 2 datagrams inside Layer 3 helps by extending VLANs across physically separated Layer 2 networks. Multiple options exist to encapsulate and tunnel protocols with varying levels of overhead, including VxLAN, NVGRE, STT, and NVO3.

The benefits of overlay networks go beyond extending VLANs. It also allows network admins to connect their physical data center to their data center in the public cloud, and to freely move their VNFs to new locations without worrying about the limitations of the underlying physical network. The network policy, security , VLANs, etc., follow the virtual machines hosting the VNFs in such a scenario, all of which make NFV deployments truly agile. But most importantly, overlay networks enable data-center-like networking on the public cloud, enabling ISVs and service providers to create virtual labs for NFV testing and PoCs on-demand using public cloud infrastructure.

In fact, recently a leading technology consulting company based out of Europe utilized AWS with overlay networking to collaborate with a service provider to build an NFV topology using Juniper Contrail , Fortinet, and F5, and use the cloud-based topology to accelerate testing for an on-premises deployment. What set this PoC apart from DC-based PoCs was that the service provider was able to model its network to scale with realistic traffic, given the unlimited capacity of AWS at its disposal.

Public Cloud + Overlay Networks = On-Demand NFVI

Public cloud coupled with overlay networking functionality creates an ideal NFVI to model full-functionality NFV topologies for feature testing and PoCs. With on-demand access to infrastructure that mimics data-center networking, along with scale to build topologies that span hundreds to thousands of nodes, service providers and network vendors alike can use public cloud to orchestrate virtual infrastructures across a wide-range of virtual and physical network, compute, and storage functions.

While still in the early stages, public-cloud based NFV infrastructure-as-a-service (IaaS) has the potential to revolutionize the networking industry much the same way AWS’s IaaS changed the course of on-premises data center investments a decade back.